Companies looking to bring breakthrough innovations to life need the right organizational structures and incentives in place. In his book, Loonshots: How to Nurture the Crazy Ideas That Win Wars, Cure Diseases, and Transform Industries (St. Martin’s Press, March 2019), Safi Bahcall, a former biotech CEO who now helps organizations innovate, explains what that requires. In this episode of the Inside the Strategy Room podcast, he talks with Erik Roth, leader of McKinsey’s innovation work globally, about why some great ideas changed the world and others failed to take off. This is an edited transcript of the discussion, which is the latest in our Committed Innovator series of conversations with innovation experts around the world. For more conversations on the strategy issues that matter, follow the series on your preferred podcast platform.
Erik Roth: You used to run a biotech company. Why did you decide to write Loonshots when you did?
Safi Bahcall: Not long after I started the biotech firm, my father was diagnosed with a rare type of leukemia. We were a cancer-focused company and I figured I could do something to help him. Unfortunately, nothing I could do made a difference and he died not long after. Over the years, as the company grew and went public, I spent a lot of time with R&D companies of all sizes and kept noticing how promising ideas, including ones that could have helped my father, remained trapped inside these organizations’ basements. That stayed with me.
Later, when I worked with President Obama’s Council of Advisors on Science and Technology [PCAST], we were tasked with looking at past national research and what we should focus on for the next century. I came across the story of how Vannevar Bush got the US military to innovate at an astonishingly fast pace and ended up changing the course of the Second World War. I thought, how could we translate what he did and how he did it into the world of large companies? That led to Loonshots.
Erik Roth: We see the same problem all the time: organizations with good intentions that cannot get out of their own way. The Vannevar Bush doctrine struck me because the military is a notoriously regimented and hierarchical organization, yet this individual was able to create agility within that system. How did you stumble on his story?
Safi Bahcall: Well, in the first meeting, the PCAST chairman said, “Your job is to write the next generation of the Vannevar Bush report.” I was a public-company CEO and had little interest in political history, so I had no idea who he was. But I found his story very interesting. Leaders of large organizations say, “Innovation is such a struggle, we’re running a 10,000-person company”—but he did it inside a two-million-person organization, and it worked. It resonated with my ideas about the importance of changing structure rather than culture. He did not change the culture. In the military, you want a consistent focus: on time, on budget, on spec. You do not want a lot of risk-taking or experiments.
Erik Roth: In our experience, many CEOs love to pull the structure lever. Unfortunately, most companies do not get Vannevar Bush results. Instead, there is a litany of failed incubators, accelerators, and other structurally separated organizations. What are they missing?
Safi Bahcall: First, we have to define structure, and you need to start not with the “what” or the “how” but the “why.” Why do I need to do these things? I showed why, as an organization grows, the incentives shift from promoting a focus on projects to a focus on politics. You can see it when the conversation around the watercooler changes from “my project” to “my career.” “What do I need to do to get promoted?” Frequently, what you need to do is the thing your boss or your boss’s boss likes—the franchise project, the thing that is easy to sell up the chain—and maybe shoot down your neighbors’ good ideas because you don’t want them succeeding.
As an organization grows, the incentives shift from promoting a focus on projects to a focus on politics. You can see it when the conversation around the watercooler changes from ‘my project’ to ‘my career.’
So, when you think about structure, you need to separate it from patterns of behavior, which is culture. If you think of water, the phase change from solid to liquid happens as you raise the temperature. You can think of culture as “are the molecules sloshing around in the glass?” That’s liquid. “Are they rigid?” That’s ice. No amount of yelling at a block of ice will melt it, but a small change in temperature can get the job done. That’s the structure.
Erik Roth: What causes the change in temperature within an organization?
Safi Bahcall: Think of reward systems: what do you incentivize? If you are rewarding intelligent risk-taking, new ideas and their outcomes, you will get an innovative culture. That is the element of structure, but it’s easy to get wrong. Many companies instead reward sounding smart in committee meetings. “That new idea on page 37, I don’t think they dotted the is and they used the wrong discount factor on page 87.” Your boss may think, “They did use the wrong discount factor on page 87. This person is really sharp, maybe we should promote them.” Boom: you reward sounding smart versus having a stake in the outcome.
Erik Roth: When you say “structure,” you are not talking boxes and lines but a broader definition.
Safi Bahcall: I mean how you design your incentive and reward systems and how you communicate them. Everything other than “Let’s get together every Friday afternoon and play ping-pong.”
Erik Roth: We have a metric we talk about: the high correlation between foosball tables and failed innovation efforts.
Safi Bahcall: The F&F correlation—foosball and failed. We used to have one at the biotech company: the shinier and bigger the building, the more likely the business is to fail.
Erik Roth: We often see a lack of a defined value, time frame, and risk level baked into companies’ strategies. We call innovation-driven net new growth “the green box.” The incentives instead simply reinforce the core business model that had long defined success. When the CEO yells at the block of ice, “Let’s go innovate,” everyone says, “It must be someone else’s job because I know what I’m supposed to be doing, it’s implicit in my reward system.” Is this why the average accelerator fails?
Safi Bahcall: Innovation labs almost always fail because they get the structure wrong. Number one, companies fail in the transfer, and number two, they fail in the incentives. You usually see a barbell structure inside companies. The operating group are the soldiers—on time, on budget, on spec consistently to customers. Then you have the creatives, on whom the business spends a ton of money.
Take the Navy: it has sailors and has thousands of research scientists at applied-physics labs working on new ideas. It is not the scientists’ priority to translate innovation into the operators’ language or understand what is working and where the hidden barriers are. Same thing with the operators: they are busy fighting wars and making sure the nuclear submarine engine doesn’t blow up. The problem the organization has not solved is the transfer.
Neither have most companies. I was meeting with the leadership team of a large investment bank and saw the same problem. Their investment bankers and private wealth managers are the best in the industry. Then the leaders say, “We need to become a technology company, but we cannot compete on hiring with start-ups, so let’s give recruits seven-figure salaries.” You get the barbell structure: a huge amount of money on a technology group, huge money on bankers and wealth managers, and who is going between the two? The high-school kid with a clipboard.
Erik Roth: This role you allude to we call the translator, the person who bridges science or technology and commercial value. Should companies turn a science person into a commercial person or make an engineering-oriented commercial person into a science person?
Safi Bahcall: The real issue is the reporting relationship. Many companies have intermediaries, but they are not taken seriously, because there is no career path. I was chatting with a military tech liaison recently who was trying to bring a new technology to the soldiers, and his counterpart’s response was, “The half-life at your job is 36 months. You are about 18 months into it, so I will just wait you out.” That is also the reality inside many companies. The liaison needs to understand the skill sets and horizontal influence and the internal product-market fit. But the intermediary’s problem is lack of prestige. One company created a separate career ladder for “program champions,” allowing them to rise all the way to the top in that role.
Let’s say Sally in the finance group comes up with an idea and shows it to Fred in the sales group, but Fred’s goal is to get Sally away from his desk as fast as possible. Why? Because he is paid on commission and every hour she is at his desk, he is not making calls. If the intermediary reports to the research group, they have no influence over Fred. If that person reports to Fred’s group, they are on Fred’s side, and it eventually escalates to the desk of the CEO, who is tearing his hair out about all the bickering.
Instead, you could establish an independent mediator group. Let’s say Melissa mediates between Sally and Fred. She speaks Fred’s language. “Fred, you don’t like these five things, but what about these other two?” Back to Sally: “Just make these few tweaks.” If one of them does not cooperate, Melissa does not escalate, because her boss isn’t either of their bosses; it is another person on the executive team. That is a change in organizational structure that shifts the problem-solving incentive to the middle. The way I describe it is that a car needs wheels and it needs pistons, but without oil, it will not run at high speeds.
Erik Roth: How should the incentives be structured?
Safi Bahcall: There is a quote attributed to Warren Buffett and Charlie Munger about incentives: “We used to say that incentives could explain 90 percent of behavior, and we were mistaken. We believe they can explain 99 percent of behavior.” If we go back to the investment-bank example, it had a barbell structure like most companies. They had a 21st-century operating group and 21st-century investment bankers and a 21st-century technology group but a 17th-century incentive system. What did they pay the investment bankers on? Transactions. What do the bankers care about? League tables. Every hour they read about some new tool is an hour they are not talking to customers.
Erik Roth: The CEO also needs to report results, and those salespeople are driving the numbers. They are incentivized, implicitly or explicitly, to leave things as they are. How do you deal with that?
Safi Bahcall: Most CEOs and boards I have spent time with want to work on something exciting— that is how they want to be remembered. The challenge is how to make that happen once you get past the 500-person mark. It’s Leadership 101: if you can’t measure it, you can’t manage it. At the same time, if the metrics are too complicated, it’s even worse. Once you can measure it, you can reward it and set up a system and a process for getting an unbiased assessment.
Erik Roth: Is there a danger of oversimplifying metrics and picking the wrong ones in these assessments?
Safi Bahcall: Yes. It’s death by NPV [net present value]. There is no better way to kill innovation than insisting, “Tell me the NPV of your project before you start.” I give the example of the transistor: probably the most important invention of the 20th century, and they could not figure out what to do with it, because it defies [American economist] Clay Christensen’s innovation theory that disruption comes from the bottom of the market. The transistor came from the largest company, not the smallest. It was expensively priced, and it was bought only by the most high-priced customers in the military.
There is no better way to kill innovation than insisting, ‘Tell me the NPV of your project before you start.’
Erik Roth: The transistor takes us into the realm of loonshots. In your book, you talk about how to nurture crazy ideas when at first the business case is not there. Is the answer to create the space and time to learn, along with the right incentives?
Safi Bahcall: I will tell you the opposite of what to do. Don’t have anybody in science who can speak business, and vice versa. Don’t make the liaison a serious job. Pay salespeople only on their output and pay science people on how cool their science stuff is. And run business-plan competitions.
Here is a typical business-plan competition: write the business case and total addressable market and calculate the NPV and the investment return rate. Two hundred people hand in 90-page documents to a three-person committee who can only pick two. What did you accomplish? You created 198 innovation anti-catalysts who are going around the watercooler saying, “This place sucks. My idea was fantastic. I am never doing this again.”
What should you do instead? A hypothesis competition: tell me your hypothesis and a way to test it in one week and with $1,000. You give people some training because hypothesis testing is an unfamiliar language for most businesspeople. Instead of “Let’s launch a new type of pizza and see if customers buy it,” you could say, “Let’s print leaflets about a new pizza flavor and put them on windshields in one afternoon and see how many people go to a website we create to get a discount.” You now have valuable data without making any pizza. With a hypothesis competition, you can fund all 200 participants; it’s worth $200,000 to create 200 innovation catalysts. Will all their experiments work? Of course not, but what is even more important is creating an experimental mindset.
Erik Roth: We all say “fail fast,” but I think what matters is learning quickly. How do you know whether an experiment was good and that you can learn from it? In your book, you suggest that people look at the wrong signals.
Safi Bahcall: Right. These are false fails. Once, when I was working on a new drug and feeling frustrated, the Nobel-laureate chemist Sir James Black, who was advising us, said, “Safi, it’s not a good drug unless it has been killed three times.” Most of the important breakthroughs failed many times before they succeeded. That is where “fail fast” goes wrong. Most companies are too impatient. There are many examples of false fails—jet engines, tanks in the military—because you can have a good product, but it needs to be married to a successful strategy. Personal music players existed long before the iPod, but Apple was good at combining product and strategy innovation.
Most of the important breakthroughs failed many times before they succeeded. That is where ‘fail fast’ goes wrong. Most companies are too impatient.
Erik Roth: How do you set up a good experiment? How do you teach a big organization to act small?
Safi Bahcall: You focus on the inputs, not the outputs. I think of it as a system mindset versus outcome mindset. What is the experimental design, and did we run the experiment well? Let’s say you launch a product and it’s a dud. The worst response is, “OK, next product.” The average response is, “Let’s have a postmortem. The product had this feature and it turns out customers preferred this versus that, so let’s do a green widget instead of a red circle next time.” That is focusing on the outcome. The great companies say, “How did we arrive at that decision? How was our system set up? How were the teams structured? What were we measuring? How did we decide to launch that product at that time with those features?
Erik Roth: Could it be that the problem they were trying to solve was not well defined? It is easy to discard a failed product because there was nothing anchoring it, forgetting that there was a market need that originally inspired the project.
Safi Bahcall: You need a system for running experiments at pace and at scale: a large number of experiments, as fast as possible, where you get high-quality data. If you define your market as too big, that is a problem. Find a group of customers with one specific job or need and one distribution channel. Are you making traction there? If not, why not? You need a clear hypothesis in advance with prespecified, numerical outcomes. If you do not succeed in your hypothesis, do you have a new one? Then you keep going.
Erik Roth: How do you help organizations not fall prey to overconfidence that they have the right answer?
Safi Bahcall: I like to do a red team/blue team loonshot exercise. The red team are the attackers. How could they kill the company? What loonshots are out there that could accomplish that? The blue team does the same but on the defense side. What loonshots are our competitors missing? What loonshots can we use to defend ourselves and overtake our competitors? You do not have the top executives on these teams; you let younger people knock themselves out with crazy ideas. This exercise liberates a lot of energy and gets people out of their comfort zone.